Counts

Counts

18 000

16 000

14 000

12 000

10 000

8 000

6 000

4 000

2 000

0

0

18 000

16 000

14 000

12 000

10 000

8 000

6 000

4 000

2 000

0

0

Cu_20kV9.95nA300s

Cu

E0 = 20 keV

500 1 000 1 500 2 000 2 500 Photon energy (eV)

Cu_20kV9.95nA300s

Residual[Cu_20kV9.95nA300s]

peak-like structure. The absorption of X-rays by the Si grid and Si dead-layer ionizes Si atoms and subsequently results in the emission of Si K-shell X-rays, which contribute a false Si peak to the spectrum. In the example for a copper target shown in .  Fig. 16.8, the apparent level of Si contributed by the internal fluorescence artifact is approximately 0.002 mass fraction.

16.2\ “Best Practices” for Electron-Excited EDS Operation

While modern EDS systems are well supported by computer automation, there remain parameters whose selection is the responsibility of the user.

16.2.1\ Operation of the EDS System

Before commencing any EDS microanalysis campaign, the analyst should follow an established checklist with careful attention to the measurement science of EDS operation. To establish the basis for quantitative analysis, the EDS parameters must be chosen consistently, especially if the analyst wishes to use archived spectra to serve as standards.

Choosing the EDS Time Constant (Resolution and Throughput)

The EDS amplifier time constant (a generic term which may be locally known as “shaping time,” “processing time,” “resolution,” “count rate range,” “1–6,” etc.) should be checked. There are usually at least two settings, one that optimizes resolution

16.2 · “Best Practices” for Electron-Excited EDS Operation

(at the cost of X-ray throughput) and one that optimizes throughput (at the cost of resolution). Confirming the desired choice of the time constant is critical for consistent recording of spectra, especially if the analyst is using archived spectra to serve as standards for quantitative analysis. This is especially important when the EDS system is in a multi-user environment, since the previous user may have altered this parameter.

Channel Width andNumber

The energy width of the histogram bins is typically chosen as 5, 10, or 20 eV. The bin energy width determines how many bins will define an X-ray peak. Since the peak width is a function of photon energy, as described by Eq. (16.2), decreasing from approximately 129 eV at Mn K-L3 (5.895 keV) to approximately 50 eV FWHM for C K-L2 (0.282 keV), a selection of a 5-eV bin width is a useful choice to optimize peak fitting since this choice will provide 10 channels across C

K-L2. The number of bins that comprise the spectrum multiplied by the bin width gives the energy span. It is useful to capture the complete energy spectrum from a threshold of 0.1 keV to the incident beam energy, E0. Thus, to span 0–20 keV with 5 eV bins requires 4096 channels.

Choosing the Solid Angle of the EDS

The solid angle Ω of a detector with an active area A at a distance r from the specimen is

If the EDS is mounted on a translatable slide that can alter the detector-to-specimen distance, then the user must select a specific value for this distance for consistency with archived standard spectra if these are to be used in quantitative analysis procedures. Because of the exponent on the distance parameter r in Eq. (16.4), a small error in r propagates to a much larger error in the solid angle and a proportional deviation in the measured intensity.

Selecting a Beam Current for an Acceptable Level of System Dead-Time

X-rays are generated randomly in time with an average rate determined by the flux of electrons striking the specimen, thus scaling with the incident beam current. As discussed above, the EDS system can measure only one X-ray photon at a time, so that it is effectively unavailable if another photon arrives while the system is “busy” measuring the first photon.

Depending on the separation in the time of arrival of the second photon, the anti-coincidence function will exclude the second photon, but if the measurement of the first photon is not sufficiently advanced, both photons will be excluded from the measurement and effectively lost. Due to this photon loss, the output count rate (OCR) in counts/second of the detector will always be less than the input count rate (ICR). The relation between the OCR and ICR is shown in .  Fig. 16.9 for a four-detector SDD-EDS. An automatic correction function measures the time increments when the detector is busy processing photons, and to compensate for possible photon loss during this “dead-time,” additional time is added at the conclusion of the user-specified measurement time so that all

. Fig. 16.9  Output count rate versus input count rate for a fourdetector SDD-EDS

Quad SDD (medium peaking time)

measurements are made on the basis of the same “live-time” so as to achieve constant dose for quantitative measurements. The level of activity of the EDS is reported to the user as a percentage “dead-time”:

Dead-time increases as the beam current increases. The dead time correction circuit can correct the measurement

time over the full dead-time range to 80 % or higher. (Note that as a component of a quality measurement system, the dead-time correction function should be periodically checked by systematically changing the beam current and comparing the measured X-ray intensity with predicted.) However, as the dead-time increases and the arrival rate of X-rays at the EDS increases, coincidence events become progressively more prominent. This effect is illustrated in

.  Fig. 16.10 for a sequence of spectra from a glass with six

Counts

16

Counts

140 000

120 000

100 000

80 000

60 000

40 000

20 000

0

0

14 000

12 000

10 000

8 000

6 000

4 000

2 000

0

0

2O K not NaK

K412

E0 = 20 keV 7% deadtime

K412 (mass frac)

K412 (mass frac)

. Fig. 16.10  Development of coincidence peaks as a function of deadtime. NIST Standard Reference Material SRM 470 (Mineral Glasses) K412. (upper) SDD-EDS spectrum at 7 % dead-time showing the characteristic

peaks for O, Mg, Al, Si, Ca, and Fe. (lower) SDD-EDS spectra recorded over arrange of dead-times showing in-growth of coincidence peaks. Note elemental misidentifications that are possible

major constituents (O, Mg, Al, Si, Ca, and Fe) measured at increasing dead-time. The spectra show the in-growth of a series of coincidence peaks as the dead-time increases. With the long pulses of the Si(Li) EDS technology, the pulse inspection function was effective in minimizing coincidence effects to dead-times in the range 20–30 %. There is more vendor-to-vendor variability in SDD-EDS technology. Some vendors provide coincidence detection that will permit dead-­times of up to 50 %, while others are restricted to 10 % dead-­time. Since there is variability among vendors’ SDD performance, it is useful to perform a measurement to determine the performance characteristic of each detector. See the sidebar for a procedure implementing such a procedure. Regardless of dead-time restrictions, an SDD-EDS is still a factor of 10 or more faster than an Si(Li)-EDS for the same resolution. In summary, as a critical step in establishing a quality measurement strategy, the beam current (for a specific EDS solid angle) should be selected to produce an acceptable rate of coincidence events in the worst-case scenario. This beam current can then be used for all measurements with reasonable expectation that the dead-time will be within acceptable limits.

kSidebar: Protocol for Determining the Optimal Probe Current and Dead-Time

Aluminum produces one of the highest fluxes of X-rays per unit probe current: With the Al K-shell ionization energy of 1.559 keV, a modest beam energy of 15 keV provides an overvoltage of 9.6 for strong excitation. Al K-L2 is of sufficient energy (1.487 keV) that it has low self-absorption, and at this energy the SDD efficiency is also relatively high. The Al K-L2 energy is low enough that this peak is also quite susceptible to coincidence events. Pure aluminum thus makes an ideal sample for testing the coincidence detection performance of a detector and for determining the maximum practical probe current for a given beam energy.

\1.\ Place a mounted, flat, polished sample of pure Al in the SEM chamber at optimal analytical working distance.

\2.\ Mount a Faraday cup with a picoammeter in the SEM chamber.

\3.\ Configure the detector at the desired process time.

\4.\ Configure the SEM at the desired beam energy and an initial probe current. Measure the probe current using the Faraday cup/picoammeter.

\5.\ Collect a spectrum from the pure Al sample with at least 10,000 counts in the Al K peak.

\6.\ Use your vendor’s software (or NIST DTSA-II) to integrate the background-corrected intensity in the Al K peak (E = 1.486 keV).

\7.\ Use your vendor’s software (or NIST DTSA-II) to look for and integrate the background-corrected intensity in the Al K + Al K coincidence peak (E = 2.972 keV).

\8.\ Determine the ratio of the integrated intensity I(Al

K + Al K)/I(Al K). We desire this ratio to be smaller than 0.01 (1 %). In some trace analysis situations, it may desirable to have this ratio less than 0.001 (0.1 %). Setting this limit too low will limit throughput but

setting it too high may make trace element analysis challenging.

\ 9.\ If the ratio is too large, decrease the probe current and re-measure the probe current and the Al spectrum. Re-measure the ratio I(Al K + Al K)/I(Al K).

\10.\ Repeat steps 5–10 until a suitable probe current has been determine.

\11.\ Finally, note the suitable probe current and use it consistently at the beam energy for which it was determined.

16.3\ Practical Aspects of Ensuring EDS

Performance for a Quality

Measurement Environment

The modern energy dispersive X-ray spectrometer is an amazing device capable of measuring the energy of tens of thousands of X-ray events per second. The spectra can be processed to extract measures of composition with a precision of a fraction of a weight-percent. However, this potential will not be realized if the detector is not performing optimally. It is important to ensure that the detector is mounted and configured optimally each time it is used. Some parameters change infrequently and need only be checked when a significant modification is made to the detector or the SEM. Other parameters and performance metrics can change from day-to-day and need to be verified more frequently. The following sections will step through a series of tests in a rationally ordered progression. The initial tests and configuration steps need only be performed occasionally, for example, when the detector is first commissioned or when a significant service event has occurred. Later steps, like ensuring proper calibration, should be performed regularly and a archival record of the results maintained.

16.3.1\ Detector Geometry

In most electron-beam instruments, the EDS detector is mounted on a fixed flange to ensure a consistent sample/ detector geometry with a fixed elevation angle. Almost all modern EDS detectors are mounted in a tubular snout with the crystal mounted at the end of the snout and the face of the active detector element perpendicular to the principle axis of the snout. The principal axis of the snout is oriented in the instrument such that it intersects with the electron beam axis at the “optimal working distance.” This geometry is illustrated in .  Fig. 16.11.

Often the detector is mounted on the flange on a sliding mechanism that allows the position of the detector to translate (move in and out) along the axis of the snout. The elevation angle is nominally held fixed during the translation but the distance from the detector crystal changes and along with it the solid angle (Ω) subtended by the detector. The solid

. Fig. 16.11  The elevation angle is a fixed property of the instrument/detector

Sample

. Fig. 16.12  The solid angle is a function of sample-to-detector distance and the active detector area

16

C

L

Objective

lens

angle is illustrated in the .  Fig. 16.12. Moving the detector away from the sample is designed to decrease the solid angle but does not change the elevation angle or the optimal working distance.

It is important to be able to maintain a reproducible solid angle through consistent repositioning of the detector. Some slide mechanisms are motorized. Motorized mechanisms should define an “inserted” position and a “retracted”

position, which you should test to ensure that the positioning is reproducible between insertions. Manual mechanisms usually provide a threaded screw with a manual crank to pull the detector in and out. The threaded screw will usually have a pair of interlocking nuts which can be positioned to define a consistent insertion position. The procedure in the sidebar below will allow you to set and maintain a constant solid angle and thus also a consistent detector collection efficiency.

zz Sidebar: Setting a Constant Detector-to-Sample Distance to Maintain Solid Angle

\1.\ Locate the pair of lock nuts on the screw mechanism. Move the lock nuts to the inner most position on the treaded rod.

\2.\ Insert the detector as close to the sample as possible. Ensure that the detector does not touch the interior of the microscope. The detector snout must be electrically isolated (no conductive path) from the interior of the chamber to eliminate noise caused by electrical ground loops.

\3.\ Twist the upper lock nut to limit the motion of the detector towards the sample. Tighten the lower nut to lock the upper nut into position.

\4.\ Test the reproducibility of the insertion point by extracting and inserting the detector and collecting a series of spectra. If the characteristic peak intensities are reproducible (to much better than a fraction of a percent) between insertions, the precision is adequate.

The take-off angle is the angle at which X-rays exit a flat sample in the direction of the detector. For a flat sample mounted perpendicular to the electron beam at the optimal working distance, the take-off angle equals the elevation angle. If the sample is tilted or the sample surface is at a slightly different working distance, then the take-off angle can be computed from the sample tilt, the working distance and the sample-to- detector distance. This is shown in .  Fig. 16.13. Often you will hear the terms elevation angle and take-off angle used interchangeably. It is more precise however to think of the elevation angle as being a fixed property of the instrument/ detector geometry and the take-off angle being dependent upon instrument-specimen configuration.

zz Check 1: Verify the Elevation Angle

It is critical that your quantitative analysis software has the correct elevation/take-off angle. Matrix correction algorithms­

use the take-off angle to calculate the correct absorption

. Fig. 16.13  The take-off angle is a function of the elevation angle, the sample tilt and the actual working distance